基于LSTM结构的文本情感分析

With the rapid development of Internet, more and more users are posting their own comments on the Internet. These comments contain a lot of valuable information, which are important for manufacturers to further understand customers\' opinions and to improve product quality. However, traditional means relying on artificial questionnaire survey are unable to meet the needs of market competition. Hence, it has been a significant research subject to extract valuable information from text ocean. In this paper, based on LSTM structure, we establish text sentiment analysis model to classify Chinese text sentiment. We also propose a pseudo gradient descent techniqueText Sentiment Analysis Based on LSTM Structure to adjust those hyper-parameters in the model. Numerical experiments show that the pseudo gradient descent technique can achieve high accuracy in shorter time.